Traditional tools for compression of image sequences are designed to encode integer data with limited dynamic range. The MPEG-4 (Moving Picture Expert Group) AVC (Advanced Video Coding)/H.264 norm (described in ISO/IEC 14496-10) and the HEVC (High Efficiency Video Coding) norm (described in B. Bross etc. JCTVC-K1003, “High Efficiency Video Coding (HEVC) text specification draft 9” October 2012) define examples of such tools.
Generally, under the AVC and HEVC norms, pixel data of a single pixel in an image comprises three components: one luminance component Y and two chrominance components U and V. Compression standards such as AVC and HEVC are used to compress images and videos whose pixel's each component is initially encoded with 8 bit depth. In extended versions of these standards, greater bit depths can be taken as input. For example, in the Range Extension version of HEVC, up to 16 bit input data is supported. The bit depth normally refers to number of bits used to indicate each component of a single pixel in an image.
Those standards are based on the discrete cosine transform (DCT). Theoretically, the coefficients obtained by performing a DCT transform on a block of an image are not necessarily integers, even though the input pixel data is an integer. In practice, however, the DCT is used for lossy compression, which means that a quantization is performed on the DCT coefficients. In this type of lossy coding scheme, it is generally preferred to use integers directly in the calculation of the DCT transform in order to reduce computation costs both at the encoder and at the decoder.
In this case, it is necessary to extend the DCT coefficients to keep sufficient precision in the calculation of DCT transform. For example, in HEVC, while input images are defined on integers with 8 or 10 bit depth, the calculation of DCT transform is performed on 32 bit depth. In the Range Extension version, up-to-16 bit depth of the input data can be supported, and the calculation of DCT transform is performed on greater bit depths, for example, 64 bit depth.